Sign up to see the future, today<\/p>\n
Can\u2019t-miss innovations from the bleeding edge of science and tech<\/p>\n
AI automation is typically exactly what it sounds like: automating tasks \u2014 many of which were previously carried out by humans \u2014 in an attempt to boost productivity and efficiency, often in a prelude to laying off workers wholesale. <\/p>\n
However, a new yet-to-be-peer-reviewed paper<\/a> conducted by a group of Microsoft researchers and spotted by IT Pro<\/a> found that today\u2019s top AI systems remain eyebrow-raisingly weak at real-world workplace tasks. In fact, they often screw them up badly: the team studied frontier models including OpenAI\u2019s GPT 5.4, Anthropic\u2019s Claude Opus 4.6 and Google\u2019s Gemini 3.1 Pro, and found that during complex assignments, those cutting edge bots corrupted an average of 25 percent of the content in documents. (Older models failed even more severely.) <\/p>\n The researchers concluded that, overall, these \u201cmodels are not ready for delegated workflows in the vast majority of domains\u201d \u2014 which is a very striking finding from Microsoft in particular, which has made massive investments in AI<\/a> and is actively trying to jam the tech<\/a> into nearly every aspect<\/a> of its Windows 11 operating system, often with disastrous results<\/a>. (Curiously, the paper didn\u2019t evaluate the company\u2019s own Copilot AI.)<\/p>\n In other words, the Redmond giant\u2019s researchers had every incentive to find something positive about AI in the workplace, but instead found that blindly trusting LLMs to handle internal documents will almost certainly result in everything from errors to data deletion.<\/p>\n As bosses everywhere push to replace human labor with AI<\/a>, the Microsoft paper builds on a growing body of scholarship about \u201cworkslop<\/a>\u201c: AI-powered mush that lazy or clueless workers push onto their colleagues, but which ultimately just needs to be fixed by a careful human laborer.<\/p>\n